Introduction

Volumetric and temporally-resolved 3-directional velocity (4D flow) MRI provides comprehensive hemodynamical analysis of blood flow through the chambers and great vessels of the heart.¹ Although whole-heart coverage can be particularly valuable for the evaluation of congenital defects characterized by pathological flow,² the extensive acquisition time required precludes its use in common clinical practice, leading to recent efforts to accelerate 4D flow acquisition.³ Furthermore, self-gating surrogate signals have been used for both respiratory and cardiac motion compensation⁴ in accelerated whole-heart 4D flow, obviating the need for ECG or time-consuming respiratory navigator-based approaches to further reduce the overhead of setup time in 4D flow imaging. In this work, we extend these concepts to propose a highly accelerated and fully self-gated 4D flow acquisition and reconstruction methodology to explore the feasibility of a 5-minute whole-heart 4D flow exam in a healthy subject.

Methods

One healthy subject was recruited for this study in accordance with the local ethics board. Highly accelerated 4D flow images were acquired with a 1.5T clinical scanner (MAGNETOM Avanto, Siemens Healthcare, Germany) for 5-minutes. Incoherent k-space sampling was achieved using a variable density, cartesian sampling pattern in the ky-kz plane (Figure 1D). Starting from the first sample at the center of the k-space, the location of each subsequent sample is advanced angularly by the golden angle and radially by another irrational number, e.g., $$$\sqrt[3](35)$$$ . The mutual irrationality of the golden angle and $$$\sqrt[3](35)$$$ leads to a k-space coverage that is approximately uniform over any arbitrary, non-contiguous time duration. A sagittal slab covering the heart and aorta was prescribed with parameters conforming to the guidelines outlined in the consensus statement (Table 1A).¹ Flow-compensated readout lines through the center of k-space were interleaved every 9 TR’s to obtain a self-gating signal (SG in Figure 1A); bandpass filtering followed by principal-component-analysis yielded self-gating surrogate signals for respiratory (Figure 1B) and cardiac (Figure 1C) motion. k-Space data were binned into 3 respiratory and 20 cardiac phases. Following binning, 4D flow images were reconstructed using the previously proposed ReVEAL4D algorithm⁵ from the respiratory expiration bin. Note that the 5-minute acquisition time included both the time to acquire the self-gating lines, which were used solely for binning, and data from respiratory phases, which were not included in the final images. Unoptimized reconstruction time for the respiratory expiration bin was approximately 6 hours (12-core Intel i7 CPU). A summary of the reconstruction pipeline is given in Figure 2. Volumetric flow rate as a function of cardiac phase, stroke volume, and peak velocity were computed in the ascending aorta. Standard breath-held 2D-phase contrast cines transecting the ascending aorta were acquired for reference Parameters for 2D-phase contrast were matched as closely as possible to the whole-heart 4D flow acquisition.

Results

Representative magnitude and phase images in the sagittal, coronal, and transverse orientations at systole are shown in Figure 3 for the reconstructed self-gated 5-minute 4D flow images. The net acceleration rate for these images is approximately 20.6. Volumetric flow rate curves as a function of cardiac phase are given in Figure 4. The self-gated 5-minute whole-heart 4D flow acquisition yields good agreement with 2D phase contrast in the ascending aorta with mean flow rates of 63.6 mL/s and 64.3 mL/s, respectively. Likewise, stroke volume and peak velocity calculated from the self-gated 5-minute whole-heart 4D flow images agree well with values derived from 2D phase contrast, with 76.4 mL and 79.3 mL, respectively for stroke volume, and 94.3 cm/s and 88.8 cm/s, respectively, for peak velocity (Table 1B).

Conclusions

In this work, we have demonstrated the feasibility of acquiring diagnostic quality whole-heart 4D flow images in 5-minutes relying only on self-gating surrogate signals for motion compensation without the need of ECG or respiratory navigators. While our proposed whole-heart 4D flow acquisition and reconstruction strategy produced good agreement with conventional 2D-phase contrast on several clinical metrics, further validation and optimization is required in a larger cohort. In the future, we will also consider implementing a “soft-gating” strategy⁶ to leverage additional information from multiple respiratory states.

Acknowledgements

This work was partially funded by NIH grants R21EB021655 and R21EB022277.